►
From YouTube: 20180808 kubeadm office hours sig cluster lifecycle
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello
today
is
August
8th
2018.
This
is
the
coop
ADM
office
hours
for
sequencer
lifecycle.
We
have
lot
of
packed
agenda,
but
we
had
stuff
on
there.
So
if
there
are
other
things
that
folks
want
to
add
to
the
agenda
for
today,
please
feel
free
to
do
so.
We
start
to
kick
it
off.
Chuck.
You
have
the
first
one,
let's
close
milestones,
unless
there's
really
just
a.
B
B
A
B
A
C
A
B
Okay,
that
sounds
great
thanks.
Tim
and
the
second
one
was:
we've
been
seeing
someone
happy
users
with
regard
to
the
changes
in
the
Debian
and
our
Debian
packages
and
RPM
packages.
Specifically,
the
problem
is
with
the
release
repo
and
how
we
changed
the
dependencies.
So
the
release
repo
builds
various
versions
of
kubernetes
using
the
one
version
using
the
master
version
of
the
release
repo,
but
we
added
the
dependencies
into
master
branch.
So
then,
any
any
new
1.8
branch,
1.9
branch
by
10
branches,
all
gonna,
start
pulling
in
those
dependencies
what
it
really
shouldn't.
B
A
Tim,
pepper
is
not
on
this
call.
I,
don't
know
what
the
state
of
basil
is
still
I,
don't
know
what,
if
there's
no
backlog,
that
I
can
track
that
I've
seen
read
somebody's
maintaining
I
think
the
problem
with
it
with
the
sig
relief
group
is
that
there
is
no
one.
That's
driving
continuity
across
the
releases
a
lot
of
the
times.
It's
a
it's
a
volunteer
group
that
once
they've
done
up
there
they're
like
peace
out,
I'm
good
all
right,
I.
Don't
necessarily
want
to
do
this
again.
A
A
So
it's
the
the
long-term
arc
of
this
was
like
back
in
1:9
was
the
original
inception,
and
we
were
only
the
only
thing.
I
knew
that
we
were
blocked
on
was
multi.
Arch
builds
for
certain
architectures
for
basil,
yeah
and
I.
Don't
know
where
that
landed,
I,
don't
know
what
the
state
of
it
is
anymore.
Thank.
A
A
So
I
don't
know,
there's
two
options:
one
is
to
fix
it
for
reals
and
two
is
to
band-aid.
For
the
time
being,
so
we
can
either
band-aid,
which
is
basically
add
the
version
checks
to
the
release
repository
or
we
can
start
to
make
a
Coalition
of
the
Willing
willing
to
fix
it
for
reals
and
try
to
actually
get
the
arc
of
basil
into
production
a
little
bit
sooner.
I'm
thinking.
Why
not
both
in
this
case,
you
could
do
both
I.
D
B
A
C
A
A
A
A
There's
precedent
and
a
lot
of
other
tooling
to
do
this
in
a
cleaner
fashion.
So
I
don't
know
if
I
want
to
just
wait
on
this
and
table
the
conversation
until
al
is
in
oh,
you
said
you
wanted
to
write
some
stuff
up
and
you
read
an
issue
for
it.
We
can
wait
till
that
point
or
everyone.
If
we're
there
more
things
to
discuss
regarding
it.
I've.
A
Think
what
I'm
going
to
try
and
do
is
I'm
going
to
try
and
think
about
this
one
because
I'm
not
happy
with
it
in
the
UX
is
weird
and
between
Ross's,
PR
and
Peter's
PR
I.
Don't
want
to
have
folks
blocked
on
this
problem,
because
they're
doing
things
slightly
differently
in
those
PRS
and
ideally
we
just
have
consistency
throughout
the
subcommands
through
to
the
config
and
make
it
abstracted
away
enough.
So
that
way
it's
just
clean
because
it
seems
like
it's.
It's
very
ad
hoc
at
this
point.
A
Two
parts
of
the
same
coin:
yes,
he
has
the
version
of
like
how
do
you
set
the
default
version
and
Rosses
PR
is
like
how
do
I
do
flag
overrides,
which
is
all
about
like
how
does
the
input
argh
and
defaults
flow
through
to
config?
So
that's
clean
throughout
the
execution
of
the
other
code
when
it's
going
through
its
phases
right.
H
F
A
The
couplet
is
really
old
too.
So,
like
don't
forget,
like
the
the
problem
with
some
of
this
stuff,
is
that
it
was
written
when
go
was
still
that
it's
four
years
old
now
and
a
lot
of
the
stuff
has
a
has
a
legacy
so,
and
some
of
the
legacy
is
folks
for
learning
go
while
they
were
building
this
stuff.
A
lot
of
them
are
old
hat
systems,
engineers
and
like
myself,
and
the
the
transition
was
not
very
clean,
so
the
best
practices
were
not
followed.
A
You
can
see
that
in
logging
and
see
that
in
command
line
argument
handling.
You
can
see
that
in
some
of
the
other
API
machinery
stuff,
you
see
that
factoring
in
the
registry.
There's
a
lot
of
areas
where
you
see
this
in
the
code
base,
so
I
don't
necessarily
know
if
I
would
look
to
existing
examples.
A
F
I'm
not
sure
if
this
is
this
idea
is
going
to
be
like
solidified,
but
so
Georgie
came.
He
came
to
the
Sikh
and
said
that
basically
they're
making
something
like
a
job
board,
but
then
the
job
idea
was
changed
to
a
community
work
because
we're
not
offering
money.
Work
is
actually
and
the
idea
is
to
have
like
a
list
of
tasks
for
users
to
be
assigned
to
like.
If
we
have
somebody
who
wants
to
work
on
documentation,
we
can
like
give
him
a
job
description
or
if
somebody
wants
to
babysit
the
end-to-end
tests.
A
A
There
was
like
15
different
cube,
ADM
ansible
installers,
and
there
were
like
dozens
of
terraform
examples
for
cube.
Idiom
it'd
be
nice
for
us
to
have
a
canonical
recipe
that
we
maintain.
So
you
know
that
we
have
as
part
of
our
testing
too,
and
that
we
can
just
say
like
here,
just
copy
this
stuff
and
go.
F
A
Google
Docs
right
now
until
they
have
like
a
template,
what's
once
they
have
a
template,
then
it
could
be
an
issue
I
guess,
but
maybe
a
work
with
the
contributes
folks
on
helping
to
get
a
template
going
makes
a
lot
of
sense
and
we
can.
We
can
revise
that
list
to
maybe
start
out
with
a
document
this
week
and
maybe
next
week
we
can
discuss
it.
F
A
We
we
can
I,
don't
it
I'm,
not
as
informed
of
what
they're
trying
to
look
for
I,
think
we
should
talk
with
Robert
and
Chris
Nova
and
ask
them
what
they
feel
they
need
from
a
Community,
Help
Wanted
sign,
I
kind
of
view.
This
as
like
you
know,
I
live
in
a
university
town,
so
there's
like
if
you
walk
down
to
the
University,
there's
like
Help
Wanted
things
all
over
the
place
or
like
little
post-it
boards,
I
kind
of
I
kind
of
envision.
It
kind
of
like
that.
F
A
I'm
not
gonna,
disagree
with
you,
I,
don't
know
whether
or
not
it's.
What
I
think
we
can
just
try
it
with
a
couple
of
simple
things
and
see
what
happens
and
then,
if
nothing
happens,
then
don't
pursue
it.
Any
further.
I
do
think
from
a
community
perspective.
The
job
we
started
in
early
110
into
111
that
we're
continuing
and
112
of
continuously
triage
in
the
backlog
is
super
helpful
for
new
contributors
that
want
to
get
engaged.
A
F
F
F
No
I
I
told
that
you're
gonna
see
something
else.
Basically,
two
men
descending.
She
was
you
can
confirm
that
I
guess
better
than
me.
She
was
away
or
perhaps
busy
with
something
else,
the
whole
the
the
whole
July
so
right
now
the
she
didn't
do
any
work
with
moving
the
dogs.
Basically,
we
have
to
move
everything
to
the
same
folder.
That
was
the
main
like
concern
that
we
have,
and
perhaps
I
should
talk
to
her
a
start.
A
A
So
I
already
mentioned
this
with
Robert,
so
he
will
hopefully
be
around
to
run
the
show
for
a
while,
but
I'll
try
to
touch
base
with
her
and
get
an
update
to
where,
if
there's
dangling
things
that
need
addressing
I
do
know
that,
within
our
repository
that
we
had
talked
about
adding
more
documentation,
we
haven't
really
done
any
of
that,
because
we
want
to
shed
some
of
the
intricacies
and
details
from
the
main,
Docs
and
just
say
like
and
go
over
to
developer
Docs.
If
you
care
about
these
details,.
F
Basically,
the
whole
process
needs
a
lot
of
decision-making,
so
we
we
still
haven't
decided
on
the
target
structure.
We
want
to
confirm.
Qadian
looks
like
the
overall
consensus
with
Lucas
in
Jennifer.
Last
time
was
it.
We
want
to
place
everything
into
a
setup
folder,
because
cube
ATM
is
mostly
set
up
related
and
the
stuff
that
is
related
to
the
CI
y
should
be
into
the
reference
phone
and
I.
If
somebody
gives
me
the
green
light,
I
can
start
merging
dogs
and
moving
stuff
around,
but
so
Jennifer
is
a
part
of
sick
dogs.
F
I
don't
have
a
green
light
from
her
because
she
wanted
to
do
some
stuff
by
himself
again
to
make
the
process
more
transparent
in
terms
of
sick
dogs
and
also
I
need
a
green/white
from
you,
like
Lumiere,
go
ahead
and
start
doing
the
work.
So
once
we
have
the
structure,
we
can
start
poking
at
the
backlog
and
implemented
stuff,
because
I
don't
want
to
have
conflicts
in
terms
of
PRS.
If
we
start
working
on
dogs
and
then
we
have
to
move
something
because
there's
a
PR
that
it's
moving
stuff
around.
Why.
A
One
thing
I
wanted
to
do
too
I
wanted
to
give
a
PSA
reminder.
I
didn't
have
in
the
notes
today
is
that
the
the
code
freeze
is
a
little
under
a
month
away.
So
that
means
any
new
feature.
Development
has
to
be
in
before
that
dates.
Only
bug
fixes
will
go
in
after
that
date,
and
I
do
know
that
there
are
some
other
work
items
that
we
wanted
to
get
in.
A
F
A
F
A
Basically
means
it's
relative,
that's
very
close,
there's
also
a
policy
for
how
you
change
things.
You
have
to
go
through
a
depreciation
cycle
and
it's
got
to
go
across
two
releases,
so
the
there
it's
actually
spelled
out.
I'd
have
to
take
up
the
docs,
but
it
is
spelled
out
with
how
you're
supposed
to
do
things
so
like
even
right
now,
I
wanted
to
get
rid
of
like
the
CLI
flag
for
self
hosting,
but
I
need
to
mark
it
as
deprecated.
First
so
yeah.
F
A
It's
not
separate
process
but
I.
Think
for
right.
Now
the
the
config
in
the
restructuring
has
been
beaten
to
death.
We
just
need
to
do
it
and
it's
it's
gotten
a
lot
of
mileage
and
thought
underneath
it
and
it
is
the
core
central
blocker
for
some
feature
enablement
like
you
mentioned,
so
we
just
need
to
get
it
done.
F
F
A
Always
try
to
use
by
default.
This
has
always
been
the
case,
and
Alexander
can
talk
about
some
of
the
history
there
that
it
always
tries
to
grab
the
latest
version
of
the
the
release
that
it
was
under
from
so
if
it's
1.11
would
be
the
late,
it
always
try
to
go
to
the
GCS
bucket
from
the
latest
version
for
that
release.
I
don't
recall
the
entire
history
for
this
decision-making
process.
I
do
remember
there
was
a
long,
be
laborious
conversation
regarding
it
due
to
like
wanting
to
have
I.
A
F
A
F
So
right
now,
I'm
suggesting
to
Peter
to
basically
use
the
version
from
the
binary,
because
Q
beta
inversion
prints
this
same
version.
Yes,
I
think
that
we
should
use
the
same
by
default
for
the
config.
Can
you
think
that
this
is
going
to
solve
the
often
problems
as
well,
and
also,
if
you,
if
we're
using
like
Q
medium
one
11.1
and
if
you
decide
to
pull
images,
is
gonna
pull
the
exact
same
image
versions,
which
I
think
is
not
a
big
problem
here?
A
B
So,
just
to
catch
you
up
to
em,
we
were
just
talking
about
the
release
process
and
how
we
have
had
a
couple
of
issues
come
in
where
we
added
some
dependencies
to
Cuba
diem
and
older
versions
like
1.8
1.9
are
being
built
with
those
dependencies
because
we're
using
the
master
branch,
the
release
repo,
and
so
we
were
wondering
like
where
the
status
of
basil
is
I.
Guess
if
you
had
any
updates
on
that.
H
Yeah,
as
best
I
could
tell
this
is
ownerless
at
the
moment.
So
I
think
that
this
stuff
is
done
sort
of
magically
behind
the
curtain
of
google
and
we've
from
the
release
team
perspective
on
any
given
release.
We
focused
on
testing
for
because,
as
we
were,
this
occur
its
actual
build
and
release
mechanism
and
how
we're
managing
the
life
cycle
on
an
ongoing
basis
and
the
older
ones.
All
of
that
has
just
sort
of
magically
happened.
H
This
is
just
building
on
the
branch
and
there
was
there
not
on
the
branch,
but
just
off
my
master
people,
then
from
Google
said
they
would
look
at
it,
but
if
you're
not
taking
ownership
of
and
it's
ambiguous,
so
the
other
gap
here
I
think
is
that
we
haven't
had
enough
volunteer
base
and
contributor
base
within
release.
Broadly,
that
has
been
the
release
team,
which
is
time
boxed
and
focused
on
the
current
really.
So
this
is
something
that
I've
been
trying
to
bring
up
at
the
cig
release
and
also
a
bit
sig
p.m.
H
B
H
F
A
We
talked
about
it
a
while
ago.
I
it's
been
well
again,
one
of
those
things
where
no
one
pulled
the
trigger,
there's,
no
one,
what
it's
kind
of
a
it's
one
of
those
things
we
said:
Mike,
wouldn't
it
be
nice
if
and
no
one's
was
there
to
follow
through,
so
we've
been
continuing
on
with
a
legacy
because
it's
because
it
works
kind
of-
and
there
hasn't
been
a
it
push
from
the
community
to
make
this
go.
I
think.
A
I
do
know
that,
like
for
a
long
time,
people
were
maintaining
their
own
medic
container
and
I
talked
about
it
within
the
sig.
For
a
long
time,
too,
I
wanted
to
get
out
of
the
business
of
actual
packaging
itself
and
to
have
like
the
made
a
container
that
does
the
installation
process,
but
that
again
you
not
enough
resources
or
time
or
volunteers
to
make
that
happen.
F
I
posted
a
comment
in
one
of
the
related
issues:
I
need
to
do
Europe.
Basically,
my
second
suggestion
here
was
basically
that
instead
of
metal
containers
like
the
go
binder
is
by
themself
a
portable
across
different
districts.
So,
in
a
way
you
don't
need
a
container
as
long
as
you
pretty
much
have
some
sort
of
an
install
script
like
the
goal.
Programming
language
binaries
are
already
portable
to
everything.
As
long
as
you
download
the
like
that
the
file
from
the
internet-
okay,.
A
I
think
if
that's
the
problem
is
UX
and
expectations
around
UX
right,
like
when
a
person
is
on
a
given
distro
they're,
very
comfortable,
with
having
a
repository
set
up
in
a
processes
in
place
for
pulling
these
things
and
using
those
things
we
even
talked
about
this
in
the
very
beginning
of
cube
ATM
and
the
the
choice
was
made
to
add
the
native
packaging
because
of
the
issues
with
UX
adoption
for
consumers.
Now,
containers
have
been
around
a
couple
years
now
and
folks
are
a
lot
more
comfortable
with
them.
A
B
As
far
as
building
go
binaries,
we
we
did
try
that
with
grad
tools
in
Covidien-
and
it
was
one
of
the
pre-flight
checks
that
failed
and
said
use,
go,
get
to
install
cried
tools
and
we
got
a
lot
of
feedback,
a
surprising
amount
of
feedback.
Actually,
that
was
saying
people
don't
want
to
install
the
go
tool
chain
to
install
a
dependency
or
installing
binary
I.
F
Guess
my
point
here:
is
it
you,
since
the
cry
tools
by
there
is
my
assumption?
Is
that
they're,
basically
portable
across
different
districts?
If
you
copy
paste
them
to
different
digital,
you
don't
have
like
like
dependencies
and
somewhere
under
the
OS
that
break
this.
So
the
battle
is
a
portable
and
you
can
use
something
a
curl
essentially
to
download
the
binary.
So
you
want
that's
like
the
most
part
approach
to
the.
C
E
A
H
F
A
F
A
That's
my
argument
for
a
long
time
you
can,
you
can
bundle
everything
into
a
cou,
medium,
specific
container
and
then,
depending
upon
the
distro,
you
might
do
some
minor
changes,
but
it
could
just
have
a
script
which
is,
like
you
know,
run
the
cuvette,
DM
install
container
run
the
install
script
for
chameleon
and
maybe
that
install
script
is
defaulted.
For
you
know
the
container
itself.
The.
E
E
A
A
We
could
talk
about
either
generic
packaging
systems
that
do
you
know,
spin-off
things
for
multiple
different
distros
talk
about
a
mini
container
and
other
options,
and
then
we
can
just
try
to
churn
on
this
topic
over
time
to
approach
consistency.
We've
done
that
with
another
number
of
other
topics,
I
think
having
the
issue
law
with
all
the
gory
details,
so
we
can
go
back
to
it
over
time
is
important.
A
A
F
Want
you
to
say
something
about
the
stock
package,
but
I
saw
this
2
micro,
Kate's
I'm,
not
sure
if
anyone
has
seen
it
it's
like
a
deployment
of
kubernetes
with
us
with
a
snap
package.
It
was
kind
of
fascinating
how
easy
it
is
to
set
up
a
question
it-it-it
executing
the
lines
of
micro,
Kate's,
dot,
Q
cattle.
E
A
All
right,
I
think
the
horse
is
sufficiently
beaten
here.
We
can
rally
on
the
issue
and
then
I
do
think.
We
should
continue
to
talk
about
it
as
part
of
next
meetings.
A
E
We
have
Docs
that
have
gone
through
the
approval
process.
You
know
the
kept
process
that
have
mastered
everywhere.
This
would
be
confusing
a
search
for
in
the
future.
We
should
also
like
retitle
the
PRS.
Are
we
allowed
to
just
like
go
in
there
and
just
say
like
hey?
Actually,
the
sake
doesn't
gonna
call
it
this
like.
Let's
bring
everything
or
do
we
have
to
like
kept
for
that
or
what
no.